Determinations of an L70 point for sodium benzoate taste detection are reported for three different psychophysical methods: Wetherill's UDTR method with single stimuli; the UDTR method modified to incorporate relative judgements at each trial; and the ascending series modification of the method of limits as commonly used in gustatory psychophysics. The high false positive rate for water stimuli interposed with sodium benzoate samples was confirmed; only the modified UDTR and the ascending series methods could be made to yield meaningful L70 determinations, which were in mutual agreement. The L70 is at about 0·04 per cent concentration; responses to sodium benzoate which have been reported below 0·01 per cent are suspect and probably mediated by an after‐taste or contrast process. Qualitative variations in the taste of sodium benzoate are reinterpreted at and below threshold.